Research map / Graph learning / GNN architectures
Interpreting Unfairness in Graph Neural Networks via Training Node Attribution
Yushun Dong, Song Wang, Jing Ma, Ninghao Liu, Jundong Li
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Delete training nodes estimated to be harmful to fairness, using a 10% deletion budget | GCN with BIND · Income | ΔSP: 25.9 $\pm$ 1.9 -> 14.7 $\pm$ 1.4 | Improved |
| Delete training nodes estimated to be harmful to fairness, using a 10% deletion budget | GCN with BIND · Pokec-z | ΔSP: 7.32 $\pm$ 2.2 -> 1.02 $\pm$ 0.9 | Improved |
| Delete training nodes estimated to be harmful to fairness, using a 10% deletion budget | GCN with BIND · Pokec-n | ΔSP: 6.57 $\pm$ 2.6 -> 2.45 $\pm$ 0.9 | Improved |
| Delete training nodes estimated to be harmful to fairness, using a 10% deletion budget | GCN with BIND · Recidivism | ΔEO: 5.23 $\pm$ 0.1 -> 4.23 $\pm$ 0.2 | Improved |
| Delete training nodes estimated to be harmful to fairness, using a 10% deletion budget | GCN with BIND · Pokec-n | ΔEO: 2.33 $\pm$ 0.5 -> 2.22 $\pm$ 1.6 | Improved |
| Delete training nodes estimated to be harmful to fairness, using a 10% deletion budget | GCN with BIND · Income | accuracy: 74.7 $\pm$ 1.4 -> 71.7 $\pm$ 0.7 | Hurt |
| Disable the non-i.i.d. loss-dependency characterization term (ablation) | BIND influence estimation · Income and other adopted datasets | Pearson correlation | Hurt |
| Randomly shuffle estimated influence values across training-node indices (ablation) | BIND influence estimation · Income | Pearson correlation | Hurt |
Closest papers by meaning
- Chasing Fairness in Graphs: A GNN Architecture Perspective (AAAI 2024 · similarity 0.90)
- Disentangling, Amplifying, and Debiasing: Learning Disentangled Representations for Fair Graph Neural Networks (AAAI 2025 · similarity 0.90)
- Theoretical and Empirical Insights into the Origins of Degree Bias in Graph Neural Networks (NeurIPS 2024 · similarity 0.90)
- Networked Inequality: Preferential Attachment Bias in Graph Neural Network Link Prediction (ICML 2024 · similarity 0.88)
- Adversarial Attacks on Fairness of Graph Neural Networks (ICLR 2024 · similarity 0.88)
- FairSIN: Achieving Fairness in Graph Neural Networks through Sensitive Information Neutralization (AAAI 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
